GALFORMOD
Semi-Analytic Models – Power to the people!
Bruno Henriques
Simon White, Gerard Lemson, Raul Angulo, Mattias Egger,
2/2/2012 Potsdam 2
GALFORMOD project
Web-based, modeler & observer friendly semi-analytic model
Combine the most robust set of dark matter numerical simulations available
Stellar Mass resolution of 108M with a large enough volume to sample BAO
MS, MII & MXXL
Munich Semi-analytic model
Scale the dark matter distribution to different cosmologies (Angulo & White 2010)
Modular implementation of the physics
“Observer friendly” outputs Choose IMF, SPS, Bands, Dust model
Monte Carlo Markov Chain optimization
2/2/2012 Potsdam 3
Outline
1 – Observer Oriented I/O
Impact of Stellar Population Synthesis Assumptions
User can select SPS, output bands, IMF and dust models
2 – Monte Carlo Markov Chain
How to find the best parameters of new models?
How to identify an unacceptable model?
How to choose between different models?
2/2/2012 Potsdam 4
Evolutionary Population Synthesis
Hot Gas
For every fraction of cold gas
mass turned into stars
Ejected Stars
Stars
Gas Knowing the metallicity and
age of the material and
Cold Gas assuming an IMF
Mass to Light
EPS assign a given spectrum
2/2/2012 Potsdam 5
Henriques et al. - 1009.1392
2/2/2012 6
Different stellar populations
Maraston, Daddi, Renzini, et al. 2006
i-band z-band K-band
2/2/2012 Potsdam 7
Reincorporation
Hot Gas
Reheating Cooling
Ejected Gas Stars
Stars Star
Recycling
Formation
Cold Gas
Ejection
2/2/2012 Potsdam 8
GALFORMOD
Dark matter simulations with a large dynamical range
Scale to different cosmologies
Observer oriented I/O – choose SPS and photometric outputs
Modular structure – easy to change existing and introduce new physics
Properties to compare/ Model Physics/
Comparison Method
Observations Parameters
MCMC
Find the region in parameter space where the chosen physics give predicted
galaxy properties that agree with observations
2/2/2012 Potsdam 9
MCMC With a Full Galaxy Catalogue
Independent trees Select a volume of the simulation where
from the dark the stellar mass function agrees with the
matter simulation total stellar mass function.
Find a representative set of merger trees,
where the galaxy properties resemble those
from the total galaxy population
Representative sample of the full
semi-analytic model in two days
( 1/512 of the Millennium volume )
Henriques, Thomas, et al. - 0810.2548 30 000 steps in 100 processors
2/2/2012 Potsdam 10
Model Physics
Radio Mode - Quiescent Black Hole Accretion Rate
kAGN=7.5x10-6
Black Hole Growth During Mergers
fBH=0.03
Cold Gas Reheating εDISK=3.5
SN Feedback Supernovae Energy εHALO=0.35
Gas Reincorporation γej=0.5
StarFormation Efficiency (αSF=0.03)
2/2/2012 Potsdam 11
Observational Constraints
set of observations that uniquely define a large number of galaxy properties
set of observations that fully constrains the parameters governing the chosen
physical processes - star formation, AGN and SN feedback
K-band Luminosity Function traces galaxy mass
Galaxy Colours star formation indicator
Black Hole - Bulge Mass relation to constrain black hole physics
2/2/2012 Potsdam 12
Observational K-band Luminosity Function
Run the semi-analytic code with a
proposed set of parameters
Compare the galaxy K-band LF with
observations getting the chi-square
probability
Accept the new parameters with that
probability and run the semi-analytic
again
The observational K-band is a combination of 3 observational data
sets - Cole et al. 2001, Bell et al. 2003, Jones et al. 2006.
With the original parameters the model overproduces the number of dwarf
galaxies.
2/2/2012 Potsdam 13
Observational K-band Luminosity Function
The contours follow the MCMC sampling in
parameter space
The colours represent the maximum likelihood
projected along the hidden dimensions
Higher ejection Less gas available to
Lower form stars in dwarfs
reincorporation Lower virial velocity cut off
Constant amount of cold gas available
2/2/2012 Potsdam 14
Ltotal=0.04
15
Tidal Disruption
Henriques & Thomas – 0909.2150
2/2/2012 Potsdam 16
Likelihood Distribution
2/2/2012 Potsdam
Model Likelihood from 0.037 to 0.15 17
Conclusions
Modular semi-analytic model, where you can change the
physics, the cosmology and the galaxy properties outputs.
Make them useful and usable by a large community - Internet
based with immediate optimization using MCMC.
Compare any set of observations and implement new
physics.
18
2/2/2012 Potsdam 19
ej mejected
m ejected
t dyn
Supernovae Feedback
Quiescent Black Hole Accretion Rate – Radio (kAGN)
m f HOT Vvir
3
kAGN=7.5x10-6
m BH , R k AGN 8 BH
10 M
0.1 200kms 1
Black Hole Growth During Mergers – Quasar (fBH)
(m sat / mcentral )mcold
m BH ,Q f BH
1 (280 kms 1 / Vvir ) 2
mreheated disk m*
fBH=0.03
εDISK=3.5 E SN 0.5 HALO m*VSN
2
Cold Gas Reheating
εHALO=0.35
Energy Released by a Supernovae ej mejected
m ejected
γej=0.5 t dyn
Gas Reincorporation
StarFormation Efficiency (αSF=0.03) ( mcold mcrit )
2/2/2012 Potsdam m SF 20
2 February, 2012 CAUP t dyn, disk 20
The Stellar Mass Function – Marchesini et al. 2009
Optical to mid-infrared data
Goods – Giavalisco et al. 2004
Musyc – Gawiser et al. 2006
2/2/2012 Potsdam 21
Uncertainties on K-corrections
Difference on the k-correction
derived using BC03 or Maraston05
2/2/2012 Potsdam 22
0.5 GYr
2/2/2012 Potsdam 23